摘要
在分析语音信号的时变自回归模型的基础上,采用了一种新的滤波器即高斯粒子滤波器,该滤波器是基于粒子滤波方法得到一高斯分布来近似估计未知状态变量的后验分布,在符合高斯假设和一定粒子数的情况下,可以获得近似最优解,并用它来解决TVAR模型的语音信号增强问题.仿真结果表明,高斯粒子滤波器具有较强的估计TVAR模型参数的能力,降低了算法的计算量.采用高斯粒子滤波增强方法处理过的语音,信噪比明显提高,改善了语音增强系统的性能.
By exploring the time-varying autoregressive models, a new Gaussian particle filter was introduced to solve speech enhancement based on the time-varying autoregressive models. A single Gaussian distribution was obtained to approximate the posterior distribution of state parameters using particle filter. GPF was asymptotically optimal with the Gaussian assumption and a certain number of particles. Simulation results show that Gaussian particle filter enhancement method not only has low computational complexity, but also improves obviously signal- to-noise ratio and the quality of speech.
出处
《江苏科技大学学报(自然科学版)》
CAS
北大核心
2009年第3期248-252,共5页
Journal of Jiangsu University of Science and Technology:Natural Science Edition
关键词
语音增强
时变自回归模型
粒子滤波器
高斯粒子滤波器
speech enhancement
time-varying autoregressive models
particle filter
Gaussian particle filter